Abstract: Automated detection and recognition of highway crack is the key to road surface crack information system.The result of crack detection will directly affect the evaluation of road damage.Based on Brawn motion model，this paper presents a multi-scale of fractal Brown motion character vector to describe image block and a k-means clustering and voting mechanism are used in produce the final result.The experimental comparisons with classic operator method（canny operator） and single-scale fractal Brown motion show that this method is more efficiency，more exactness and closer to original crack position in detecting road crack weak information.Furthermore，the method takes on very high practice value.

Key words: road crack, fractal Brown motion, multi-scale, weak information

摘要： 路面裂缝的自动检测与识别是路面病害信息系统的关键环节，裂缝检测结果的优劣直接影响路面损坏程度的评估。基于布朗运动模型，采用不同尺度下的分数布朗运动特征参数矢量描述图像块特征，并运用k聚类与投票机制产生最终结果。通过与经典算子（canny）及单一尺度布朗运动矢量的比较结果证明，该算法在检测路面破损弱信息时更有效，检出的结果更准确、更接近原始裂缝位置，具有实际应用价值。

关键词: 路面破损, 分数布朗运动, 多尺度, 弱信息